Given a set of correlated securities and their corresponding closing prices, I am looking for a way to construct an index of these securities.

When applying PCA to obtain the principal components, how can I use or understand the output to construct an index?

For instance in matlab:

[coeff,score,latent,tsquared,explained,mu] = pca(prices)
  • $\begingroup$ Not sure I understood what you are trying to do. The eigenvector i from the PCA is in fact the weights of an index that would track the -ith principal component. $\endgroup$
    – Alex C
    Apr 24 '17 at 0:21
  • 1
    $\begingroup$ Hmm clear. But if you choose lets say the first two PC to explain the variation, how do you derive then the weight for the j-th asset? $\endgroup$ Apr 24 '17 at 0:24
  • $\begingroup$ I don't know. I would form a linear combination of the first two PC, weighted by the square roots of the respective eigenvalues (so the volatilities are properly taken into account). $\endgroup$
    – Alex C
    Apr 24 '17 at 0:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.